{ "id": 679336, "title": "Massive bot attack", "url": "/competitions/deep-past-initiative-machine-translation/discussion/679336", "author_name": "", "post_date": "2026-02-28T18:04:45.880091600Z", "votes": 13, "comment_count": 18, "views": 0, "content": "

Today I noticed an unusual spike in the number of teams entering the competition. I took a look at the public LB and found out that more than 140 solo participants joined the competition and submitted the same public notebook. It is usually 15-20 teams a day for the last 2-3 weeks. Places 1272–1412 are now occupied by freshly created contributor accounts with no history.

\n

Does anyone know what it is? Are these bots? Or did some institution ask students to participate? Nicknames are diverse and submission intervals are even 4 to 5 minutes — a clear uniform distribution! I do think these are bots. I have never seen anything like this.

\n

@ryanholbrook can you please take a look. It looks very suspicious

\n

EDIT: I took my investigation further and manualy checked arbitrary 10 accounts. They ALL joined exactly DPC, Stanford RNA 3D Folding Part 2 and AIMO. And ALL of them made exactly 1 submission

\n

EDIT2: Now these accounts submitted more notebooks. Most of them now have 3 submissions. And now they do not occupy 140+ consecutive places and it's not easy to detect all of them in few minutes.

\n

\"\"

", "messages": [ { "id": "3415341", "postDate": "02/28/2026 18:04:45", "content": "

Today I noticed an unusual spike in the number of teams entering the competition. I took a look at the public LB and found out that more than 140 solo participants joined the competition and submitted the same public notebook. It is usually 15-20 teams a day for the last 2-3 weeks. Places 1272–1412 are now occupied by freshly created contributor accounts with no history.

\n

Does anyone know what it is? Are these bots? Or did some institution ask students to participate? Nicknames are diverse and submission intervals are even 4 to 5 minutes — a clear uniform distribution! I do think these are bots. I have never seen anything like this.

\n

@ryanholbrook can you please take a look. It looks very suspicious

\n

EDIT: I took my investigation further and manualy checked arbitrary 10 accounts. They ALL joined exactly DPC, Stanford RNA 3D Folding Part 2 and AIMO. And ALL of them made exactly 1 submission

\n

EDIT2: Now these accounts submitted more notebooks. Most of them now have 3 submissions. And now they do not occupy 140+ consecutive places and it's not easy to detect all of them in few minutes.

\n

\"\"

", "rawMarkdown": "Today I noticed an unusual spike in the number of teams entering the competition. I took a look at the public LB and found out that **more than 140 solo participants** joined the competition and submitted the same public notebook. It is usually 15-20 teams a day for the last 2-3 weeks. Places 1272–1412 are now occupied by freshly created contributor accounts with no history.\n\nDoes anyone know what it is? Are these bots? Or did some institution ask students to participate? Nicknames are diverse and submission intervals are even 4 to 5 minutes — **a clear uniform distribution**! I do think these are bots. I have never seen anything like this.\n\n@ryanholbrook can you please take a look. It looks very suspicious\n\nEDIT: I took my investigation further and manualy checked arbitrary 10 accounts. They **ALL** joined **exactly** DPC, Stanford RNA 3D Folding Part 2 and AIMO. And **ALL of them made exactly 1 submission**\n\nEDIT2: Now these accounts submitted more notebooks. Most of them now have 3 submissions. And now they do not occupy 140+ consecutive places and it's not easy to detect all of them in few minutes. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13977249%2Fb9cd7d4593cd5d45773a56f54ada3bc6%2Fweird.jpg?generation=1772301535489616&alt=media)", "votes": null }, { "id": "3415346", "postDate": "02/28/2026 18:18:07", "content": "

Probably bots, or at least someone trying to probe to better understand the test set now that it should be final. Or maybe there's a classroom that decided to join, but it's probably the first one.

", "rawMarkdown": "Probably bots, or at least someone trying to probe to better understand the test set now that it should be final. Or maybe there's a classroom that decided to join, but it's probably the first one.", "votes": null }, { "id": "3415347", "postDate": "02/28/2026 18:18:56", "content": "

Thats not a probing, because they all have submitted the same notebook. May be a preparation for it. And a classroom version doesnt seem valid as well because of weird submission intervals — a clear uniform distribution

", "rawMarkdown": "Thats not a probing, because they all have submitted the same notebook. May be a preparation for it. And a classroom version doesnt seem valid as well because of weird submission intervals — a clear uniform distribution", "votes": null }, { "id": "3415351", "postDate": "02/28/2026 18:23:49", "content": "

Ah, I missed that part. I assumed that it was someone sending the same thing with just a small change each time (gap spacing, hyphens, etc.) to better understand for themselves the fuzzier rules that people are debating. But maybe not.

\n

Edit: thank you for having noticed it and flagged it.

", "rawMarkdown": "Ah, I missed that part. I assumed that it was someone sending the same thing with just a small change each time (gap spacing, hyphens, etc.) to better understand for themselves the fuzzier rules that people are debating. But maybe not.\n\nEdit: thank you for having noticed it and flagged it.", "votes": null }, { "id": "3415352", "postDate": "02/28/2026 18:25:04", "content": "

Nope. Because they are sorted by a submission time! When 2 notebooks have the same score, earlier submission is higher on the LB

", "rawMarkdown": "Nope. Because they are sorted by a submission time! When 2 notebooks have the same score, earlier submission is higher on the LB", "votes": null }, { "id": "3415370", "postDate": "02/28/2026 19:31:19", "content": "

My guess is they’re farming submissions on AIMO but trying to not stand out… in a pretty bad way though 😅

", "rawMarkdown": "My guess is they’re farming submissions on AIMO but trying to not stand out… in a pretty bad way though 😅", "votes": null }, { "id": "3415373", "postDate": "02/28/2026 19:41:44", "content": "

Yea, could be. But why did they made only 1 sub on AIMO as well?

", "rawMarkdown": "Yea, could be. But why did they made only 1 sub on AIMO as well?", "votes": null }, { "id": "3415385", "postDate": "02/28/2026 20:15:37", "content": "

Limit one per day. That comp is currently running a lottery approach as well… the top scores are from public notebooks that have massive variance in scores run to run

", "rawMarkdown": "Limit one per day. That comp is currently running a lottery approach as well… the top scores are from public notebooks that have massive variance in scores run to run", "votes": null }, { "id": "3415387", "postDate": "02/28/2026 20:17:14", "content": "

I know. But they submitted much earlier in AIMO. Like 1-2 weeks ago. So there should be much more submission if lottery win is an aim

", "rawMarkdown": "I know. But they submitted much earlier in AIMO. Like 1-2 weeks ago. So there should be much more submission if lottery win is an aim", "votes": null }, { "id": "3415688", "postDate": "03/01/2026 06:27:31", "content": "

My guess: it has to be those medal sellers again, they are trying to artificially increase the team count so that there will be more medals available and hence a higher chance to win one

", "rawMarkdown": "My guess: it has to be those medal sellers again, they are trying to artificially increase the team count so that there will be more medals available and hence a higher chance to win one", "votes": null }, { "id": "3415744", "postDate": "03/01/2026 08:53:56", "content": "

Interesting find. If this is a lottery strategy for AIMO as suggested, wouldn't 140 unique seeds on a public baseline be the most cost-effective way to secure a medal in a high-variance environment? How can the platform differentiate these from legitimate newcomers without compromising the validation anchor?

", "rawMarkdown": "Interesting find. If this is a lottery strategy for AIMO as suggested, wouldn't 140 unique seeds on a public baseline be the most cost-effective way to secure a medal in a high-variance environment? How can the platform differentiate these from legitimate newcomers without compromising the validation anchor?", "votes": null }, { "id": "3415747", "postDate": "03/01/2026 09:00:48", "content": "

Do you agree that 1400 unique seeds is a better strategy? Because they submitted in AIMO 1-2 weeks ago once and never submitted again. And every new submission is a new seed basically

", "rawMarkdown": "Do you agree that 1400 unique seeds is a better strategy? Because they submitted in AIMO 1-2 weeks ago once and never submitted again. And every new submission is a new seed basically", "votes": null }, { "id": "3415766", "postDate": "03/01/2026 09:29:17", "content": "

I had no idea that artificially inflating team counts to unlock more medal slots was even a thing. Has anything like this been observed in previous competitions? Would be curious to know if Kaggle has dealt with this before and how it was resolved.

", "rawMarkdown": "I had no idea that artificially inflating team counts to unlock more medal slots was even a thing. Has anything like this been observed in previous competitions? Would be curious to know if Kaggle has dealt with this before and how it was resolved.", "votes": null }, { "id": "3416068", "postDate": "03/02/2026 01:47:46", "content": "

openclaw ??

", "rawMarkdown": "openclaw ??", "votes": null }, { "id": "3416114", "postDate": "03/02/2026 05:20:47", "content": "

I think some people just want to increase the number of submissions to increase the number of medals, thereby raising their own chances of getting a medal.

\n

But I think the more likely situation is that there’s a “lottery” going on in AIMO and RNA prediction.

\n

To be honest, I also “lottery” twice a day in RNA, and the lottery results score much higher than my own method. Of course I know that’s just the LB, and those lottery results won’t carry over to the private LB—haha.

", "rawMarkdown": "I think some people just want to increase the number of submissions to increase the number of medals, thereby raising their own chances of getting a medal.\n\nBut I think the more likely situation is that there’s a “lottery” going on in AIMO and RNA prediction.\n\nTo be honest, I also “lottery” twice a day in RNA, and the lottery results score much higher than my own method. Of course I know that’s just the LB, and those lottery results won’t carry over to the private LB—haha.", "votes": null }, { "id": "3416117", "postDate": "03/02/2026 05:25:12", "content": "

Also, submissions from accounts like that usually end up getting banned by the organizers sooner or later. That’s probably why the leaderboard can still change even a few days after the competition ends.\nIn Santa’s, there was one unlucky guy who got hit by this—he dropped from Silver to Bronze.

", "rawMarkdown": "Also, submissions from accounts like that usually end up getting banned by the organizers sooner or later. That’s probably why the leaderboard can still change even a few days after the competition ends.\nIn Santa’s, there was one unlucky guy who got hit by this—he dropped from Silver to Bronze.", "votes": null }, { "id": "3416285", "postDate": "03/02/2026 14:52:03", "content": "

This makes more sense than my initial thought. I had no idea there were medal sellers but I guess there's a market for everything

", "rawMarkdown": "This makes more sense than my initial thought. I had no idea there were medal sellers but I guess there's a market for everything", "votes": null }, { "id": "3416293", "postDate": "03/02/2026 15:07:31", "content": "

Hi @samson8,

\n

Thanks for the heads up. I've alerted our moderation team. They will be investigating.

", "rawMarkdown": "Hi @samson8,\n\nThanks for the heads up. I've alerted our moderation team. They will be investigating.", "votes": null }, { "id": "3421346", "postDate": "03/15/2026 10:20:07", "content": "

maybe i think, or oyher agent

", "rawMarkdown": "maybe i think, or oyher agent", "votes": null } ], "comments": [ { "id": 3415346, "author_name": "epiktistes", "author_url": "", "post_date": "02/28/2026 18:18:07", "content": "

Probably bots, or at least someone trying to probe to better understand the test set now that it should be final. Or maybe there's a classroom that decided to join, but it's probably the first one.

", "votes": null, "replies": [ { "id": 3415347, "author_name": "samson8", "author_url": "", "post_date": "02/28/2026 18:18:56", "content": "

Thats not a probing, because they all have submitted the same notebook. May be a preparation for it. And a classroom version doesnt seem valid as well because of weird submission intervals — a clear uniform distribution

", "votes": null, "replies": [ { "id": 3415351, "author_name": "epiktistes", "author_url": "", "post_date": "02/28/2026 18:23:49", "content": "

Ah, I missed that part. I assumed that it was someone sending the same thing with just a small change each time (gap spacing, hyphens, etc.) to better understand for themselves the fuzzier rules that people are debating. But maybe not.

\n

Edit: thank you for having noticed it and flagged it.

", "votes": null, "replies": [ { "id": 3415352, "author_name": "samson8", "author_url": "", "post_date": "02/28/2026 18:25:04", "content": "

Nope. Because they are sorted by a submission time! When 2 notebooks have the same score, earlier submission is higher on the LB

", "votes": null, "replies": [] } ] } ] } ] }, { "id": 3415370, "author_name": "jackvd", "author_url": "", "post_date": "02/28/2026 19:31:19", "content": "

My guess is they’re farming submissions on AIMO but trying to not stand out… in a pretty bad way though 😅

", "votes": null, "replies": [ { "id": 3415373, "author_name": "samson8", "author_url": "", "post_date": "02/28/2026 19:41:44", "content": "

Yea, could be. But why did they made only 1 sub on AIMO as well?

", "votes": null, "replies": [ { "id": 3415385, "author_name": "jackvd", "author_url": "", "post_date": "02/28/2026 20:15:37", "content": "

Limit one per day. That comp is currently running a lottery approach as well… the top scores are from public notebooks that have massive variance in scores run to run

", "votes": null, "replies": [ { "id": 3415387, "author_name": "samson8", "author_url": "", "post_date": "02/28/2026 20:17:14", "content": "

I know. But they submitted much earlier in AIMO. Like 1-2 weeks ago. So there should be much more submission if lottery win is an aim

", "votes": null, "replies": [] } ] } ] } ] }, { "id": 3415688, "author_name": "yeoyunsianggeremie", "author_url": "", "post_date": "03/01/2026 06:27:31", "content": "

My guess: it has to be those medal sellers again, they are trying to artificially increase the team count so that there will be more medals available and hence a higher chance to win one

", "votes": null, "replies": [ { "id": 3415766, "author_name": "nkforwork", "author_url": "", "post_date": "03/01/2026 09:29:17", "content": "

I had no idea that artificially inflating team counts to unlock more medal slots was even a thing. Has anything like this been observed in previous competitions? Would be curious to know if Kaggle has dealt with this before and how it was resolved.

", "votes": null, "replies": [] }, { "id": 3416285, "author_name": "jackvd", "author_url": "", "post_date": "03/02/2026 14:52:03", "content": "

This makes more sense than my initial thought. I had no idea there were medal sellers but I guess there's a market for everything

", "votes": null, "replies": [] } ] }, { "id": 3415744, "author_name": "oreo01mofu", "author_url": "", "post_date": "03/01/2026 08:53:56", "content": "

Interesting find. If this is a lottery strategy for AIMO as suggested, wouldn't 140 unique seeds on a public baseline be the most cost-effective way to secure a medal in a high-variance environment? How can the platform differentiate these from legitimate newcomers without compromising the validation anchor?

", "votes": null, "replies": [ { "id": 3415747, "author_name": "samson8", "author_url": "", "post_date": "03/01/2026 09:00:48", "content": "

Do you agree that 1400 unique seeds is a better strategy? Because they submitted in AIMO 1-2 weeks ago once and never submitted again. And every new submission is a new seed basically

", "votes": null, "replies": [] } ] }, { "id": 3416068, "author_name": "hengzheng", "author_url": "", "post_date": "03/02/2026 01:47:46", "content": "

openclaw ??

", "votes": null, "replies": [ { "id": 3421346, "author_name": "chenleonid", "author_url": "", "post_date": "03/15/2026 10:20:07", "content": "

maybe i think, or oyher agent

", "votes": null, "replies": [] } ] }, { "id": 3416114, "author_name": "ngyzly", "author_url": "", "post_date": "03/02/2026 05:20:47", "content": "

I think some people just want to increase the number of submissions to increase the number of medals, thereby raising their own chances of getting a medal.

\n

But I think the more likely situation is that there’s a “lottery” going on in AIMO and RNA prediction.

\n

To be honest, I also “lottery” twice a day in RNA, and the lottery results score much higher than my own method. Of course I know that’s just the LB, and those lottery results won’t carry over to the private LB—haha.

", "votes": null, "replies": [ { "id": 3416117, "author_name": "ngyzly", "author_url": "", "post_date": "03/02/2026 05:25:12", "content": "

Also, submissions from accounts like that usually end up getting banned by the organizers sooner or later. That’s probably why the leaderboard can still change even a few days after the competition ends.\nIn Santa’s, there was one unlucky guy who got hit by this—he dropped from Silver to Bronze.

", "votes": null, "replies": [] } ] }, { "id": 3416293, "author_name": "ryanholbrook", "author_url": "", "post_date": "03/02/2026 15:07:31", "content": "

Hi @samson8,

\n

Thanks for the heads up. I've alerted our moderation team. They will be investigating.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3415341": "Today I noticed an unusual spike in the number of teams entering the competition. I took a look at the public LB and found out that **more than 140 solo participants** joined the competition and submitted the same public notebook. It is usually 15-20 teams a day for the last 2-3 weeks. Places 1272–1412 are now occupied by freshly created contributor accounts with no history.\n\nDoes anyone know what it is? Are these bots? Or did some institution ask students to participate? Nicknames are diverse and submission intervals are even 4 to 5 minutes — **a clear uniform distribution**! I do think these are bots. I have never seen anything like this.\n\n@ryanholbrook can you please take a look. It looks very suspicious\n\nEDIT: I took my investigation further and manualy checked arbitrary 10 accounts. They **ALL** joined **exactly** DPC, Stanford RNA 3D Folding Part 2 and AIMO. And **ALL of them made exactly 1 submission**\n\nEDIT2: Now these accounts submitted more notebooks. Most of them now have 3 submissions. And now they do not occupy 140+ consecutive places and it's not easy to detect all of them in few minutes. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13977249%2Fb9cd7d4593cd5d45773a56f54ada3bc6%2Fweird.jpg?generation=1772301535489616&alt=media)", "3415346": "Probably bots, or at least someone trying to probe to better understand the test set now that it should be final. Or maybe there's a classroom that decided to join, but it's probably the first one.", "3415347": "Thats not a probing, because they all have submitted the same notebook. May be a preparation for it. And a classroom version doesnt seem valid as well because of weird submission intervals — a clear uniform distribution", "3415351": "Ah, I missed that part. I assumed that it was someone sending the same thing with just a small change each time (gap spacing, hyphens, etc.) to better understand for themselves the fuzzier rules that people are debating. But maybe not.\n\nEdit: thank you for having noticed it and flagged it.", "3415352": "Nope. Because they are sorted by a submission time! When 2 notebooks have the same score, earlier submission is higher on the LB", "3415370": "My guess is they’re farming submissions on AIMO but trying to not stand out… in a pretty bad way though 😅", "3415373": "Yea, could be. But why did they made only 1 sub on AIMO as well?", "3415385": "Limit one per day. That comp is currently running a lottery approach as well… the top scores are from public notebooks that have massive variance in scores run to run", "3415387": "I know. But they submitted much earlier in AIMO. Like 1-2 weeks ago. So there should be much more submission if lottery win is an aim", "3415688": "My guess: it has to be those medal sellers again, they are trying to artificially increase the team count so that there will be more medals available and hence a higher chance to win one", "3415744": "Interesting find. If this is a lottery strategy for AIMO as suggested, wouldn't 140 unique seeds on a public baseline be the most cost-effective way to secure a medal in a high-variance environment? How can the platform differentiate these from legitimate newcomers without compromising the validation anchor?", "3415747": "Do you agree that 1400 unique seeds is a better strategy? Because they submitted in AIMO 1-2 weeks ago once and never submitted again. And every new submission is a new seed basically", "3415766": "I had no idea that artificially inflating team counts to unlock more medal slots was even a thing. Has anything like this been observed in previous competitions? Would be curious to know if Kaggle has dealt with this before and how it was resolved.", "3416068": "openclaw ??", "3416114": "I think some people just want to increase the number of submissions to increase the number of medals, thereby raising their own chances of getting a medal.\n\nBut I think the more likely situation is that there’s a “lottery” going on in AIMO and RNA prediction.\n\nTo be honest, I also “lottery” twice a day in RNA, and the lottery results score much higher than my own method. Of course I know that’s just the LB, and those lottery results won’t carry over to the private LB—haha.", "3416117": "Also, submissions from accounts like that usually end up getting banned by the organizers sooner or later. That’s probably why the leaderboard can still change even a few days after the competition ends.\nIn Santa’s, there was one unlucky guy who got hit by this—he dropped from Silver to Bronze.", "3416285": "This makes more sense than my initial thought. I had no idea there were medal sellers but I guess there's a market for everything", "3416293": "Hi @samson8,\n\nThanks for the heads up. I've alerted our moderation team. They will be investigating.", "3421346": "maybe i think, or oyher agent" }, "source": "meta" }